added sentiment/emoji labels
Browse files- data/tweet_hate/map.txt +8 -0
- data/tweet_sentiment/test.jsonl +0 -0
- data/tweet_sentiment/train.jsonl +0 -0
- data/tweet_sentiment/validation.jsonl +0 -0
- process/tweet_sentiment.py +13 -0
- super_tweet_eval.py +8 -7
data/tweet_hate/map.txt
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
0,hate_gender
|
2 |
+
1,hate_race
|
3 |
+
2,hate_sexuality
|
4 |
+
3,hate_religion
|
5 |
+
4,hate_origin
|
6 |
+
5,hate_disability
|
7 |
+
6,hate_age
|
8 |
+
7,not_hate
|
data/tweet_sentiment/test.jsonl
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
data/tweet_sentiment/train.jsonl
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
data/tweet_sentiment/validation.jsonl
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
process/tweet_sentiment.py
CHANGED
@@ -59,6 +59,19 @@ train['text'] = train['text'].apply(clean_text)
|
|
59 |
validation['text'] = validation['text'].apply(clean_text)
|
60 |
test['text'] = test['text'].apply(clean_text)
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
# save splits
|
63 |
cols_to_keep = ['gold_label', 'topic', 'text']
|
64 |
train[cols_to_keep].to_json(
|
|
|
59 |
validation['text'] = validation['text'].apply(clean_text)
|
60 |
test['text'] = test['text'].apply(clean_text)
|
61 |
|
62 |
+
# map classes to 0-4
|
63 |
+
class_map = {
|
64 |
+
-2:0,
|
65 |
+
-1:1,
|
66 |
+
0:2,
|
67 |
+
1:3,
|
68 |
+
2:4
|
69 |
+
}
|
70 |
+
train['gold_label'] = train['gold_label'].map(class_map)
|
71 |
+
validation['gold_label'] = validation['gold_label'].map(class_map)
|
72 |
+
test['gold_label'] = test['gold_label'].map(class_map)
|
73 |
+
|
74 |
+
|
75 |
# save splits
|
76 |
cols_to_keep = ['gold_label', 'topic', 'text']
|
77 |
train[cols_to_keep].to_json(
|
super_tweet_eval.py
CHANGED
@@ -2,7 +2,7 @@
|
|
2 |
import json
|
3 |
import datasets
|
4 |
|
5 |
-
_VERSION = "0.1.
|
6 |
_SUPER_TWEET_EVAL_CITATION = """TBA"""
|
7 |
_SUPER_TWEET_EVAL_DESCRIPTION = """TBA"""
|
8 |
_TWEET_TOPIC_DESCRIPTION = """
|
@@ -274,8 +274,6 @@ class SuperTweetEval(datasets.GeneratorBasedBuilder):
|
|
274 |
features["gold_score"] = datasets.Value("float32")
|
275 |
if self.config.name == "tempo_wic":
|
276 |
features["gold_label_binary"] = datasets.Value("int32")
|
277 |
-
# features["token_idx_1"] = datasets.Value("int32")
|
278 |
-
# features["token_idx_2"] = datasets.Value("int32")
|
279 |
features["text_start_1"] = datasets.Value("int32")
|
280 |
features["text_start_2"] = datasets.Value("int32")
|
281 |
features["text_end_1"] = datasets.Value("int32")
|
@@ -285,9 +283,8 @@ class SuperTweetEval(datasets.GeneratorBasedBuilder):
|
|
285 |
if self.config.name == "tweet_hate":
|
286 |
label_classes = [
|
287 |
'hate_gender','hate_race', 'hate_sexuality', 'hate_religion','hate_origin', 'hate_disability',
|
288 |
-
'
|
289 |
features['gold_label'] = datasets.features.ClassLabel(names=label_classes)
|
290 |
-
#features["gold_label"] = datasets.Value("int32")
|
291 |
features["text"] = datasets.Value("string")
|
292 |
if self.config.name == "tweet_nerd":
|
293 |
features['target'] = datasets.Value("string")
|
@@ -297,10 +294,14 @@ class SuperTweetEval(datasets.GeneratorBasedBuilder):
|
|
297 |
features['text_end'] = datasets.Value("int32")
|
298 |
features['gold_label_binary'] = datasets.Value("int32")
|
299 |
if self.config.name == "tweet_emoji":
|
300 |
-
|
|
|
|
|
|
|
301 |
features["text"] = datasets.Value("string")
|
302 |
if self.config.name == "tweet_sentiment":
|
303 |
-
|
|
|
304 |
features["text"] = datasets.Value("string")
|
305 |
features["target"] = datasets.Value("string")
|
306 |
|
|
|
2 |
import json
|
3 |
import datasets
|
4 |
|
5 |
+
_VERSION = "0.1.38"
|
6 |
_SUPER_TWEET_EVAL_CITATION = """TBA"""
|
7 |
_SUPER_TWEET_EVAL_DESCRIPTION = """TBA"""
|
8 |
_TWEET_TOPIC_DESCRIPTION = """
|
|
|
274 |
features["gold_score"] = datasets.Value("float32")
|
275 |
if self.config.name == "tempo_wic":
|
276 |
features["gold_label_binary"] = datasets.Value("int32")
|
|
|
|
|
277 |
features["text_start_1"] = datasets.Value("int32")
|
278 |
features["text_start_2"] = datasets.Value("int32")
|
279 |
features["text_end_1"] = datasets.Value("int32")
|
|
|
283 |
if self.config.name == "tweet_hate":
|
284 |
label_classes = [
|
285 |
'hate_gender','hate_race', 'hate_sexuality', 'hate_religion','hate_origin', 'hate_disability',
|
286 |
+
'hate_age', 'not_hate']
|
287 |
features['gold_label'] = datasets.features.ClassLabel(names=label_classes)
|
|
|
288 |
features["text"] = datasets.Value("string")
|
289 |
if self.config.name == "tweet_nerd":
|
290 |
features['target'] = datasets.Value("string")
|
|
|
294 |
features['text_end'] = datasets.Value("int32")
|
295 |
features['gold_label_binary'] = datasets.Value("int32")
|
296 |
if self.config.name == "tweet_emoji":
|
297 |
+
with open('./data/tweet_emoji/map.txt') as f:
|
298 |
+
label_classes = f.readlines()
|
299 |
+
label_classes = [x.strip('\n') for x in label_classes]
|
300 |
+
features['gold_label'] = datasets.features.ClassLabel(names=label_classes)
|
301 |
features["text"] = datasets.Value("string")
|
302 |
if self.config.name == "tweet_sentiment":
|
303 |
+
label_classes = ["strongly negative", "negative", "negative or neutral", "positive", "strongly positive"]
|
304 |
+
features['gold_label'] = datasets.features.ClassLabel(names=label_classes)
|
305 |
features["text"] = datasets.Value("string")
|
306 |
features["target"] = datasets.Value("string")
|
307 |
|